UQCRFS1N assembles mitochondrial respiratory complex-III into an asymmetric 21-subunit dimer

Mitochondrial respiratory chain consists of four multimeric protein complexes, Complex I-IV (CI, NADH dehydrogenase; CII, succinate:ubiquinone oxidoreductase; CIII, cytochrome bc1 complex; and CIV, cytochrome c oxidase). These four complexes transfer electrons from NADH or FADH2 to oxygen and pump protons from mitochondrial matrix to intermembrane space, generating electrochemical gradient across the inner membrane which is harnessed by complex V to synthesize ATP, providing the majority of energy acquired by living organisms. Respiratory chain complexes were reported to interact with each other to form supercomplexes, even megacomplex (Guo et al., 2017). However, despite decades of intensive research, many intriguing new findings concerning respiratory chain still emerging frequently in recent years (Baradaran et al., 2013; Guo et al., 2016; Schagger and Pfeiffer, 2000; Vinothkumar et al., 2014). CIII is the middle part of respiratory chain, transferring electrons from ubiquinone to cytochrome c, and pumping protons in the meantime. All previous studies reported that CIII from vertebrate species was a homodimer consisting of 22 subunits in total (Fernandez-Vizarra and Zeviani, 2015, 2017), while after analysis of previous structural data, we draw a different conclusion. The structures of the CIII homodimer from different species have been extensively studied, with the first structure of bovine CIII solved in 1997 by the Deisenhofer group (Xia et al., 1997). Since then, a total of 46 (23 bovine, 17 chicken, 2 porcine, 3 ovine, and 1 human) CIII structures from vertebrate species are available till now (Fernandez-Vizarra and Zeviani, 2017). In all these structures, CIII was descripted as a homodimer. However, unlike other subunits, the UQCRFS1N subunit in all these structures is poorly defined. Although several structures have been solved with very high resolution, the full length N-terminal processed peptide (1– 78 amino acids, UQCRFS1N) of the iron-sulfur Rieske protein (UQCRFS1) subunit has not been assigned in all of these structures (Table 1). UQCRFS1N is the N-terminal mitochondrial targeting sequence of UQCRFS1, and after its cleavage from the precursor, this small peptide remains bound to CIII with unknown functions. In this letter, we show that one UQCRFS1N links the two 10-subunit CIII protomers together to form the intact CIII, which resultantly contains only 21 subunits rather than previously assumed 22 subunits (Fig. 1A and 1B). Firstly, we rebuilt the high-resolution crystal structures of bovine CIII (PDB: 2A06) (Huang et al., 2005) and chicken CIII (PDB:3TGU) (Hao et al., 2012). To do this, we downloaded the bovine structure (PDB: 2A06), the chicken structure (PDB:3TGU) and their corresponding MTZ files. We deleted UQCRFS1N residues in those two models and refined them with their maps thus generating two meshes named mFo–DFc and 2mFo–DFc, colored with green and blue, respectively. Analysis was processed after modelling of UQCRFS1N back into the corresponding densities. Intriguingly, the 2.1 Å density map of bovine CIII clearly shows that it contains two opposite orientations of the fulllength UQCRFS1N peptide (1–78 amino acids), with the N terminal binding in one CIII protomer and the C terminal binding in the other CIII protomer (Fig. 1C–E). Both orientations can fit very well with the refined map, where the terminals of the two orientations occupy different densities while the middle regions of the two orientations share the same density. Refinement of the highest resolution chicken structure (PDB:3TGU) (Hao et al., 2012) gave same results (Fig. 1F).


Dear Editor,
Mitochondrial respiratory chain consists of four multimeric protein complexes, Complex I-IV (CI, NADH dehydrogenase; CII, succinate:ubiquinone oxidoreductase; CIII, cytochrome bc 1 complex; and CIV, cytochrome c oxidase). These four complexes transfer electrons from NADH or FADH 2 to oxygen and pump protons from mitochondrial matrix to intermembrane space, generating electrochemical gradient across the inner membrane which is harnessed by complex V to synthesize ATP, providing the majority of energy acquired by living organisms. Respiratory chain complexes were reported to interact with each other to form supercomplexes, even megacomplex (Guo et al., 2017). However, despite decades of intensive research, many intriguing new findings concerning respiratory chain still emerging frequently in recent years (Baradaran et al., 2013;Guo et al., 2016;Schagger and Pfeiffer, 2000;Vinothkumar et al., 2014). CIII is the middle part of respiratory chain, transferring electrons from ubiquinone to cytochrome c, and pumping protons in the meantime. All previous studies reported that CIII from vertebrate species was a homodimer consisting of 22 subunits in total Zeviani, 2015, 2017), while after analysis of previous structural data, we draw a different conclusion.
The structures of the CIII homodimer from different species have been extensively studied, with the first structure of bovine CIII solved in 1997 by the Deisenhofer group (Xia et al., 1997). Since then, a total of 46 (23 bovine, 17 chicken, 2 porcine, 3 ovine, and 1 human) CIII structures from vertebrate species are available till now (Fernandez-Vizarra and Zeviani, 2017). In all these structures, CIII was descripted as a homodimer. However, unlike other subunits, the UQCRFS1N subunit in all these structures is poorly defined.
Although several structures have been solved with very high resolution, the full length N-terminal processed peptide (1-78 amino acids, UQCRFS1N) of the iron-sulfur Rieske protein (UQCRFS1) subunit has not been assigned in all of these structures (Table 1). UQCRFS1N is the N-terminal mitochondrial targeting sequence of UQCRFS1, and after its cleavage from the precursor, this small peptide remains bound to CIII with unknown functions. In this letter, we show that one UQCRFS1N links the two 10-subunit CIII protomers together to form the intact CIII, which resultantly contains only 21 subunits rather than previously assumed 22 subunits ( Fig. 1A and 1B).
Firstly, we rebuilt the high-resolution crystal structures of bovine CIII (PDB: 2A06) (Huang et al., 2005) and chicken CIII (PDB:3TGU) (Hao et al., 2012). To do this, we downloaded the bovine structure (PDB: 2A06), the chicken structure (PDB:3TGU) and their corresponding MTZ files. We deleted UQCRFS1N residues in those two models and refined them with their maps thus generating two meshes named mFo-DFc and 2mFo-DFc, colored with green and blue, respectively. Analysis was processed after modelling of UQCRFS1N back into the corresponding densities. Intriguingly, the 2.1 Å density map of bovine CIII clearly shows that it contains two opposite orientations of the fulllength UQCRFS1N peptide (1-78 amino acids), with the N terminal binding in one CIII protomer and the C terminal binding in the other CIII protomer ( Fig. 1C-E). Both orientations can fit very well with the refined map, where the terminals of the two orientations occupy different densities while the middle regions of the two orientations share the same density. Refinement of the highest resolution chicken structure (PDB:3TGU) (Hao et al., 2012) gave same results (Fig. 1F).

LETTER TO THE EDITOR
This composition is conserved in vertebrates. Unconventionally, we propose that a single UQCRFS1N molecule tethers the two 10-subuint CIII protomers, with the N-terminal segment (residues 1-36) bound to one protomer and the C-terminal segment (residues 42-78) bound to the other. Furthermore, from the extra densities, we could build another UQCRFS1N molecule into CIII in a switched orientation, with the N-and C-terminal segments of the peptide inserting into the opposite CIII protomers ( Fig. 1A and 1B). The two UQCRFS1N molecules partially overlap with each other, indicating that they could not simultaneously exist in the same complex. Thus, they likely represent alternative conformations of CIII. Thus, each CIII complex has 21 subunits. Twenty subunits (each protomer containing ten subunits) form a symmetric dimer while the UQCRFS1N peptide inserts into the two protomers with two alternative orientations to assemble them together ( Fig. 1A and 1B).
The unusual binding mode of the UQCRFS1N peptide in CIII has not been previously observed. We checked 29 mammalian CIII (23 from bovine, 2 from porcine, 3 from ovine and 1 from human) and 17 chicken CIII structures available in the RCSB PDB protein data bank (https://www. rcsb.org), and found that none of them contains the fulllength UQCRFS1N peptide (Table 1), consistent with a previous report (Fernandez-Vizarra and Zeviani, 2017). All previous structural studies on CIII suggested that this complex was a symmetric homodimer. However, our refinement of bovine and chicken structures (PDB: 2A06 and 3TGU) indicates that a single UQCRFS1N peptide can insert into the two CIII protomers with two different orientations when CIII is assembled into the respirasome (Fig. 1A and 1B). This further suggests that CIII has no preferred orientation when it binds to CI and CIV to form the respirasome Wu et al., 2016). Re-refinement of the previous high-resolution X-ray bovine and chicken CIII structures (PDB: 2A06 and 3TGU) suggests that a single UQCRFS1N peptide in two opposite orientations can tether the two CIII protomers together ( Fig. 1A and 1B), even in the absence of CI and CIV. This phenomenon is further consistent with our recent findings that both protomers of CIII have the abilities to bind with CI to form the megacomplex-I 2 III 2 IV 2 (Guo et al., 2017).
The full length UQCRFS1 is the last subunit to be incorporated into CIII. Previous studies suggested that the tetratricopeptide repeat domain 19 (TTC19) protein collaborates with an inner membrane protease SLP2-PARL-YME1L (SPY) complex to remove UQCRFS1N (residues 1-78 amino acids) from the full length UQCRFS1 after its incorporation into CIII (Bottani et al., 2017;Wai et al., 2016). In the mitochondria of the Ttc19 -/knockout mouse, the native molecular mass of CIII was slightly higher than that of CIII isolated from the wild type animals. This small difference might be the incorporation of two UQCRFS1N segments (which are still parts of the full-length UQCRFS1) into CIII, as opposed to the one UQCRFS1N subunit in the mature CIII. More recently, it was suggested that the CIII core subunits, UQCRC1 and UQCRC2, have the matrix processing peptidase (MPP) activity to cleave the UQCRFS1N peptide after UQCRFS1 is incorporated into the nascent pre-CIII dimer (Fernandez-Vizarra and Zeviani, 2017). On the other hand, UQCRFS1N could inhibit the MPP activity of the UQCRC1-  (Deng et al., 2001). Our structures show that one UQCRFS1N molecule can bind to both cavities formed by UQCRC1 and UQCRC2 in both protomers. It is unclear whether UQCRC1-UQCRC2 is indeed involved in the processing of UQCRFS1. More biological experiments are needed to clarify how UQCRFS1 is processed and how CIII is assembled.
In conclusion, after refinement of previous high resolution crystal structures of CIII, we find UQCRFS1N can incorporate into CIII dimer in two possible orientations, indicating CIII is a 21-subunit asymmetric dimer rather than a 22-subunit homodimer, and both orientations of UQCRFS1N can exist in respirasome, supporting that both protomers of CIII in respirasome are functional and are possible to interact with CI to form the megacomplex I 2 III 2 IV 2 .

LETTER TO THE EDITOR
Shuai Zong et al.